Multi-Channel Sparsity Histogram based Particle Filter for Hand Tracking

Sparse representation has been widely applied for modeling object appearance in the tracking process. Various sparsity models are coded by single feature with fixed local patches, which neglects the potential spatial information inside the appearance model. This way also gives a low robust effect on appearance variations such as partial occlusion, illumination and scale variation. In order to resolve above problems, a novel sparsity representation based on multi-channel features is presented in this paper. Pure color pixel vectors in the local patch are decomposed into several separate dictionaries, which can keep the structural attributes of original model appearance. After constructing sparsity histogram in the local patches, a cosine histogram measurement is proposed for the comparison between template and candidate from particle filter framework. Finally, a scheme for template update by incremented 2DPCA learning is employed for appearance variation. In order to show the proposed method robust performance, this work employs hand as tracking target that traditional skin Color based methods which are easily effected by lots of factors and difficult to track the hand target. Qualitative experimental result shows that the proposed tracking algorithm gives a better performance in dynamic scenes.